Abstract

This paper analyses the applicability of artificial neural networks for predicting the lubricated friction coefficient. We will consider their use as faster and simpler alternatives to simulations based on theoretical behaviour equations. The development of several different artificial neural networks is presented. They have been trained through tribological tests on a mini-traction-machine, which furnishes the friction coefficient in point contacts. Once the training has been completed the networks are applied as tools for predicting the results in different operating conditions. Their advantages and disadvantages are analysed compared with conventional simulation tools. Copyright (c) 2013 John Wiley Sons, Ltd.